Targeted Molecular Profiling of Circulating Cell-Free DNA in Patients With Advanced Hepatocellular Carcinoma

Authors

Cowzer D

White JB

Chou JF

Chen PJ

Kim TH

Khalil DN

El Dika IH

Columna K

Yaqubie A

Light JS

Shia J

Yarmohammadi H

Erinjeri JP

Wei AC

Jarnagin W

Do RKG

Solit DB

Capanu M

Shah RH

Berger MF

Abou-Alfa GK

Harding JJ

Doi

10.1200/PO.23.00272

PMID: 37769223 · DOI: 10.1200/PO.23.00272 · Journal: JCO Precision Oncology (2023)

TL;DR

This single-center retrospective study at Memorial Sloan Kettering evaluated targeted NGS-based molecular profiling of circulating cell-free DNA (cfDNA) using the 129-gene MSK-ACCESS assay in 51 patients with histologically confirmed advanced hepatocellular carcinoma. Genomic alterations were detected in 92.2% of patients. The most frequently mutated genes were TERT promoter (57%), TP53 (47%), CTNNB1 (37%), ARID1A (18%), and TSC2 (14%). Plasma-tissue concordance was high for matched samples (n=37), but 27% of paired samples harbored cfDNA-exclusive alterations, including potentially actionable mutations, supporting complementary use of both liquid and tissue-based profiling.

Cohort & data

  • 53 plasma samples from 51 patients with histologically confirmed advanced HCC at MSKCC (February 2019 to August 2021).
  • All patients had inoperable disease or were ineligible for liver transplant; AJCC stage IV in 74%, stage III in 18%, stage II in 8%.
  • Median age 69 years (range 42-87); 76% male; 47% hepatitis B/C etiology, 53% non-viral.
  • 47% were treatment-naive at time of cfDNA collection; 35% had prior TKI; 18% prior anti-PD-1/PD-L1; 8% both.
  • cfDNA profiled with MSK-ACCESS (129-gene panel, ~200,000x raw coverage with error suppression); matched tumor tissue profiled with MSK-IMPACT in 37 (72.5%) patients.
  • Variant annotation via OncoKB.
  • Data deposited in cBioPortal as hcc_jcopo_msk_2023.

Key findings

  • Detection rate: Genomic alterations detected in 48/53 (90.6%) plasma samples and 47/51 (92.2%) patients. Median cfDNA yield 39.43 ng (range 7.93-287.68); median VAF 0.027 (range 0.001-0.28).
  • Mutation frequencies in cfDNA: TERT promoter 57%, TP53 47%, CTNNB1 37%, ARID1A 18%, TSC2 14%, APC 8%, CDKN2A 8%, PIK3CA 8%, RB1 6%, BRCA1 4%, BRCA2 4%, NF1 4%, TSC1 4%.
  • Pathway enrichment: WNT-beta-catenin pathway altered in 45% of cases; PI3K-AKT-mTOR pathway in 25%. No microsatellite instability detected.
  • Plasma-tissue concordance (n=37 paired): cfDNA detected 92.5% of alterations previously called in matched tumor tissue. Per-gene concordance: TERT 83% (kappa 0.67), TP53 94% (kappa 0.74), CTNNB1 92% (kappa 0.83), ARID1A 100% (kappa 0.91), TSC2 71% (kappa 0.65), PIK3CA 50% (kappa 0.36).
  • cfDNA-exclusive alterations: In 10/37 (27%) paired samples, cfDNA detected mutations not found in matched tumor tissue, including TSC2, CTNNB1, PIK3CA, BRCA2, KRAS, TERT, ARID1A, and TP53. Of these, 4/10 (40%) were OncoKB level 3b actionable.
  • Comparison to published tissue cohort (n=121): cfDNA mutation frequencies were not significantly different from tissue for TERT (61% vs 56%, P=0.6), CTNNB1 (41% vs 36%, P=0.6), ARID1A (18% vs 14%, P=0.5), and TSC2 (16% vs 9%, P=0.2). TP53 was significantly more prevalent in cfDNA (51% vs 32%, P=0.024).
  • VAF associations (multivariable): Higher average VAF associated with AFP >=400 ng/mL (beta 0.43, P=0.007), higher total tumor volume (beta 0.14, P<0.001), and no previous systemic therapy (beta -0.52, P<0.001).
  • Survival: In treatment-naive patients (n=24), cfDNA VAF (high vs low, dichotomized at median) was not associated with overall survival (HR 0.88, 95% CI 0.28-2.75, P=0.82).

Genes & alterations

Gene Alteration type Frequency (cfDNA) Key finding
TERT Promoter mutations 57% Most frequent; 83% concordance with tissue
TP53 Missense, nonsense, frameshift 47% 94% concordance; significantly more frequent in cfDNA vs tissue (P=0.024)
CTNNB1 Missense, multihit 37% 92% concordance; WNT pathway driver
ARID1A Missense, frameshift, nonsense 18% 100% concordance (kappa 0.91)
TSC2 Splice site, frameshift 14% Actionable (OncoKB 3b); two patients received everolimus
TSC1 Various 4% Combined TSC1/2 alterations in 18%
PIK3CA Oncogenic 8% Actionable (OncoKB 3b); PI3K-AKT-mTOR pathway
BRCA1 Various 4% Combined BRCA1/2 in 8%; actionable
BRCA2 Various 4% cfDNA-exclusive in one paired sample
APC Various 8% WNT pathway
CDKN2A Various 8% Cell-cycle pathway
RB1 Various 6% Cell-cycle pathway
NF1 Various 4% RAS-MAPK pathway
KRAS Various — cfDNA-exclusive in one paired sample (OncoKB level 4)

Clinical implications

  • cfDNA profiling via MSK-ACCESS is a viable alternative to tissue-based genomic profiling in advanced HCC, particularly when tissue acquisition is not feasible due to cirrhosis or other contraindications.
  • Potentially actionable mutations were identified in 37% of patients, including TSC1/TSC2 (18%), BRCA1/BRCA2 (8%), and PIK3CA (8%), supporting cfDNA-guided enrollment in basket or disease-specific clinical trials.
  • Two patients received everolimus based on mTOR-AKT pathway alterations identified in tissue and confirmed by cfDNA; one achieved stable disease for 5.4 months (TSC2 splice site mutation), the other progressed after 2.5 months (TSC2 frameshift deletion).
  • Complementary use of both cfDNA and tissue sequencing may be necessary to detect all actionable alterations, as 27% of paired samples showed cfDNA-exclusive mutations (40% of which were actionable).
  • Higher VAF correlates with disease burden (AFP, tumor volume) and treatment-naive status, suggesting improved analytical yield in the pre-treatment setting.

Limitations & open questions

  • Single-center retrospective study at MSKCC with potential selection bias; only 51 patients.
  • Heterogeneous timing of cfDNA collection relative to systemic therapy (47% treatment-naive, 41% post-TKI, 18% post-immunotherapy).
  • Small sample size for survival analysis in treatment-naive subset (n=24); VAF did not correlate with OS but the study was underpowered for this endpoint.
  • Median time between tumor and cfDNA sample collection was 9.8 months (range 0.1-70.6), introducing potential for clonal evolution to explain discordant findings.
  • MSK-ACCESS covers 129 genes; AXIN1 and BAP1 (recurrently altered in HCC tissue studies) were not covered by the panel, limiting the landscape assessment.
  • Only one exon each of JAK1 and ARID2 was covered by MSK-ACCESS, potentially underestimating their prevalence.
  • Prospective validation in larger cohorts is needed to establish the prognostic significance of cfDNA detection and VAF in advanced HCC.

Citations from this paper used in the wiki

  • “Forty-eight (90.6%) of 53 plasma samples had detectable genomic alterations, with a positive sample detected in 47 of 51 (92.2%) patients.” PMID:37769223
  • “The most frequently mutated genes were TERT promoter (57%), TP53 (47%), CTNNB1 (37%), ARID1A (18%), and TSC2 (14%).” PMID:37769223
  • “Plasma cfDNA sequencing identified 92.5% of genomic alterations that had been reported clinically on tissue sequencing.” PMID:37769223
  • “In 10 (27%) of 37 tumor-plasma samples, alterations were detected by cfDNA analysis that were not detected in the patient-matched tumors.” PMID:37769223
  • “Clinically actionable mutations were identified in cfDNA in 37% of cases, including TSC1/2 (18%), BRCA1/2 (8%), and PIK3CA (8%).” PMID:37769223
  • “Higher average VAF was associated with AFP >=400 ng/mL (beta .43; CI, 0.12 to 0.74; P=.007), higher tumor volume (beta .14; CI, 0.07 to 0.22; P<=.001), and no previous therapy (beta -.52; CI, -0.83 to 0.21; P<=.001).” PMID:37769223

This page was processed by crosslinker on 2026-05-04.